Stuck at fault model

• Source this playlist on Testing.

Introduction

• Fault model applied at the logic level
• Allows us to design test vectors
• Allows us to determine wether the test pattern is sufficient to detect all the possible faults
• Allows us to project fault coverage statistics

Model

• It looks at every single node of the circuit and says that said node could have three state
  • normal
  • stuck at logic zero
  • stuck at logic one
• In the example node F is stuck at logic zero

• When we say that a node in a logic circuit is stuck at zero or one we aren’t actually implying anything about the underlying defect that caused the fault
  • node F stuck at zero doesn’t mean that F is shorted to ground
• It’s just a way to represent defects, and it could represent alot of defects, and it doesn’t imply anything about actual connections that happen in the finished chips

Example

• OR gate with 2 inputs and 1 output, It has N=3 nodes
• Number of faults that can occur = K*N where K is the number of faults that can occur at every node
• In stuck at fault model we have 2 faults per node K=2
  • six possible faults that can occur for this circuit
• We will assume that there is only one stuck-at fault happening
  • depends on the fact that the probability of a single fault occuring is much higher than the probability of multiple faults happen
• The table contains the normal output, and the output when each of the faults occur
• This table allows us to see which inputs can be used to detect which faults
  • This can be done be comparing each of the columns of the table with the correct output of the circuit
• Applying inputs A=0,B=0 isn’t useful in trying to identify the presence of F stuck at zero because the correct output matches the output when stuck
  • on the other hand applying any of the other three inputs of the truth table is gonna expose F stuck at zero because the output deviates from the true output in all three cases
  • This is called an error because it’s an observation that deviates from the true output and it exposes an underlying fault
• A stuck at zero will only be exposed when applying the inputs A=1,B=0
• It’s useful to look at the minimum set of test vectors which can expose all the possible faults that can occur in the circuit
  • This is not necessarily the entirety of the truth table
• When we try to get the minimum set that exposes all the faults we start by faults that are uncovered by only a single test F fault at one then we start to move to faults that are uncovered by more
• In the example test vector T={00,10,01} we don’t need to apply input 11 because the only fault exposed by applying input 11 is F stuck at zero which is also exposed by applying {01,10}
• Distinguishing which fault occured through the test vector is possible in some cases and isn’t possible in others
  • If we applied test vector 01 and a fault occured this can be due to F stuck at zero or B stuck at one, this can be known by applying test 10 if a fault happens then it’s F stuck at zero, if no fault happens then it’s B stuck at one
  • Faults F, A, B stuck at one have the same observation so they can’t be distinguished if a fault happens